Be on the lookout around campus today for a troupe of Alex Ganns on the loose this Halloween Cold Spring Harbor Laboratory @wsbsbot

New work led by Shushan Toneyan and Ziqi (Amber) Tang (Amber Tang) on evaluating deep learning models for predicting epigenomic profiles. #RegulatoryGenomics 1/5 biorxiv.org/content/10.110…

Let me preach to you the way of W&B…

First in person conference at #ASHG22 ! Come chat with me at poster PB2983 today if you want to chat about our work evaluating deep learning for predicting epigenomic profiles!

Excited to share new work from my lab on "Evolution-inspired augmentations improve deep learning for regulatory genomics" Paper: biorxiv.org/content/10.110… Code: github.com/p-koo/evoaug Analysis: github.com/p-koo/evoaug_a… 1/N

Excited to share new work with ShToneyan! CREME (Cis-Regulatory Element Model Explanations), a suite of in silico perturbation experiments to uncover the rules of gene regulation learned by large-scale DNNs trained on functional genomics data. biorxiv.org/content/10.110…

Excited to share new work on "Interpreting cis-regulatory mechanisms from genomic deep neural networks using surrogate models” led by Evan Seitz, jointly advised by me and Justin B. Kinney and in collab with David M. McCandlish Paper: biorxiv.org/content/10.110… Docs: squid-nn.readthedocs.io

Excited for #MLCB2023! Check out talk by ShToneyan on uncovering higher-order CRE interactions from large-scale DNNs and 2 posters: revisiting inits by Chandana Rajesh and Amber Tang and new attribution method using domain-inspired surrogate models by Evan Seitz! biorxiv.org/content/10.110…

Do current genomic language models (pre-trained on whole genomes) learn a foundational understanding of biology in the non-coding region of human genomes? A new evaluation led by Amber Tang suggests not yet! 1/N paper: biorxiv.org/content/10.110…

Very excited to be at SysBio2024. Shoot me a message or just come chat with me about deep learning in regulatory genomics! #cshlsysbio

In genomic deep learning, the trends right now are to build bigger models that consider longer sequence contexts. While predictions are more powerful, their scale makes them difficult to interpret. To address this gap, we have developed CREME. Paper: biorxiv.org/content/10.110… 1/N

Excited to sponsor and participate in the Machine Learning for Computational Biology Conference #MLCB2024, kicking off today in Seattle! Watch Thomas Pierrot explain how InstaDeep is advancing Generative AI for Genomics! 🧬 📽️ Live here: bit.ly/47hkcVb

Excited to share HIPPO (Histopathology Interventions of Patches for Predictive Outcomes)! HIPPO is a perturbation-based post hoc explanation tool interprets weakly supervised models for digital pathology. 1/N Work led by @JakubKaczmarzyk.

🧬 Genomic DNNs can be trained to learn a lot of different aspects of gene regulation, but they're not perfect and we don't know which predictions are reliable and which ones aren't. We introduce DEGU: Uncertainty-aware Genomic Deep Learning with Knowledge Distillation. 1/n

Thrilled to open-source the dataset behind our Nature Machine Intelligence cover paper! 🧬 The ChatNT training data is now open-source on Hugging Face. It's the first large-scale dataset for training conversational agents on biological sequences. A thread on what's inside 👇

Our work on "Evaluating the representational power of pre-trained DNA language models for regulatory genomics" led by Amber Tang with help from Nirali Somia & Steven Yu is finally published in Genome Biology! Check it out! genomebiology.biomedcentral.com/articles/10.11…